Identifying mergers using non-parametric morphological classification at high redshifts

TL;DR

This study evaluates the effectiveness of non-parametric morphological indicators, especially asymmetry, in identifying major galaxy mergers at high redshifts ($z=2$ to 4) using cosmological simulations.

Contribution

It demonstrates that asymmetry is the most reliable morphological indicator for detecting major mergers at high redshifts, with specific thresholds improving detection accuracy.

Findings

High asymmetry values correlate strongly with recent major mergers.

Canonical asymmetry cut (A≥0.35) detects only about 10% of mergers.

Higher asymmetry thresholds (A≥0.8 or 1.5) improve merger identification.

Abstract

We investigate the time evolution of non-parametric morphological quantities and their relationship to major mergers between $4\geq z \geq 2$ in high-resolution cosmological zoom simulations of disk galaxies that implement kinetic wind feedback, $H_2$-based star formation, and minimal ISM pressurisation. We show that the resulting galaxies broadly match basic observed physical properties of $z\sim 2$ objects. We measure the galaxies' concentrations ($C$), asymmetries ($A$), and $Gini$ ($G$) and $M_{20}$ coefficients, and correlate these with major merger events identified from the mass growth history. We find that high values of asymmetry provide the best indicator for identifying major mergers of $>1:4$ mass ratio within our sample, with $Gini$-$M_{20}\,$ merger classification only as effective for face-on systems and much less effective for edge-on or randomly-oriented galaxies. The canonical asymmetry cut of $A\geq0.35$, however, is only able to correctly identify major mergers $\sim 10\%$ of the time, while a higher cut of $A\geq 0.8$ more efficiently picks out mergers at this epoch. We further examine the temporal correlation between morphological statistics and mergers, and show that for randomly-oriented galaxies, half the galaxies with $A\geq0.8$ undergo a merger within $\pm0.2\,{\rm Gyr}$, whereas $Gini$-$M_{20}\,$ identification only identifies about a third correctly. The fraction improves further using $A\geq 1.5$, but about the half the mergers are missed by this stringent cut.